The Four Necessary but Insufficient Conditions for a Bubble | A Conversation with Economist Professor 朱宁
Summary
朱宁 does not condemn bubbles ex ante, but identifies four “necessary but insufficient conditions”: a new concept or technology, easy liquidity, government support, and young investors who lack experience but are willing to bet. AI broadly meets these conditions, but that does not mean it must crash, or even that it will break in the near term; the actionable conclusion is not to guess tomorrow’s outcome, but to ask: “Given that I don’t know the future, what is the right investment decision and choice for me today?”
The most dangerous near-term signal at a bubble’s peak is often not more shorts, but the last shorts being crushed by the rally and flipping long. Tiger Fund liquidated in 2000, and the Nasdaq peaked 1-2 months later; 巴鲁克 regarded the moment when beggars, shoeshine boys, and nail technicians started teaching people how to trade stocks as a top signal. Once “everyone who thought this was a bubble has been crushed by reality,” the last leg of buying may be nearly complete.
Investors should focus on controlling position size and risk exposure, not obsess over whether AI stocks will be higher or lower 6 months from now. 朱宁 advocates participating first through diversification and partial exposure, then planning separately for what to do after a rally and after a selloff; “you cannot expect to make all the money in the world,” and making an extra 5 points or 10 points matters far less than avoiding one major drawdown.
Long-term returns are shaped first by the era’s Beta, while personal Alpha still determines whether the result can endure, but understanding determines whether money made through luck can be kept. 巴菲特 has acknowledged the “ovarian dividend” of being born in the United States and receiving an education, while 95% of his wealth was accumulated after age 65; 利弗莫尔 became rich shorting in 1929, then blew up through overconfidence and leverage, illustrating that “you can never make money beyond your understanding.”
AI will push the internet era’s democratization of information toward “democratization of capability,” but it may also expose retail investors to both model hallucinations and self-hallucinations. Institutions still retain advantages in research, IT, organizational systems, and discipline, while a market structure in which indexes rise even as more individual stocks fall will further reinforce index investing; when 朱宁 asked GPT and DeepSeek what to do, the answer was: “If you can’t beat them, join them.”
AI is a great technology, and some foundation-model companies can genuinely make money, but that does not mean AI stocks deserve their current valuations. 朱宁 recalled that 席勒’s framework put U.S. equity valuations around 2018 second only to 2000; he also said current U.S. valuations are roughly 20% above 1929 levels. 曲凯 added that it remains unclear whether downstream companies calling the models will be able to make money.
Monetary policy can delay, smooth, or shift an adjustment, but it may simply use one bubble to pay for another. Post-dot-com easing helped create a housing bubble, which could then be followed by another internet or AI narrative; low rates force capital into risk, speculation, and leverage, while rates and funding stress may eventually become “the last straw that breaks the camel’s back.”
A small principal base is not automatically a reason to go all in; young people should weigh financial investment alongside human-capital investment. Those who can repeatedly go to zero without regret may be able to earn money through their risk tolerance, but most people want to take risks without being able or willing to bear the losses; 朱宁’s bottom line is that “we make money to live, not live to make money,” and the more valuable advantage of youth is the ability to trade small losses for experience, fall down, and start again.
Deep dive
1. Bubbles Repeat Human Nature Before They Repeat Asset Names
朱宁 traces the behavioral roots of bubbles to 3 forces: overconfidence, linear extrapolation, and herding. As prices rise, investors treat the outcome as proof of forecasting ability—“See? I told you the stock would go up”—and then become even more convinced they can see the future because they got that call right.
Linear thinking is both the simplest and most dangerous form of inference: the market rose yesterday, so it will “probably keep rising tomorrow, right? Why? Because it rose yesterday.” The logic is irrational, but it fits the mental habits humans developed through evolution, which is why it reappears in every era.
Herding eliminates the need for research. If you do not understand stocks or property but enough people around you are participating and making money, you can persuade yourself that joining is the right decision. 朱宁 closed with 黑格尔’s line: “The only thing we learn from history is that we learn nothing from history.”
2. When a Bubble Does Not Burst, It Is Called “Irrational Exuberance”
曲凯 described the classic latecomer’s path: recognizing a technology’s value early, questioning the gains in the middle phase and staying on the sidelines, then entering out of fear of missing out—only to find that the market seems to break just as they arrive. The same pattern could emerge in AI because an attractive value proposition and significant price risk can coexist.
朱宁 follows his mentor 席勒’s framework from the 2000 book Irrational Exuberance: “As long as this bubble does not burst, it is irrational exuberance; after it bursts, we call it a bubble in hindsight.” No one can therefore prove responsibly in advance that a given boom must be a bubble.
That uncertainty is not a reason to abandon decision-making; it changes the question. Investors should not force themselves to settle the philosophical issue of whether something is “really a bubble” before deciding to buy everything or nothing. The secondary market allows them to choose how much to invest while uncertainty remains.
3. Design the Response to Up and Down Markets Before Deciding Whether to Buy
朱宁 rejects framing an investment decision as either all in or completely out. Regardless of how bullish or bearish the view is, an investor can start with a partial position: a rally will not trigger panic over missing out, while a selloff will leave losses manageable. Diversification keeps emotions from being held hostage by a single direction.
Late-cycle chasing often carries a compensation motive: “I missed the move, so now I have to make up for the time and money I lost.” That impulse is not driven by new information, but by the emotional response to others making money while you did not—and it is exactly what can push investors into a value trap.
巴菲特 once made money trading cocoa futures, but later emphasized that someone in the world would certainly make money trading cocoa or chocolate futures; “but that is not my investment style.” 朱宁 distilled the boundary: “You cannot expect to make all the money in the world.”
The more useful question for AI stocks is not whether they will crash tomorrow or 6 months from now, but what the additional return from a rally would mean for you, and whether a loss would damage your life and long-term plans. 曲凯 relayed the same answer from a secondary-market investor: do not bet on direction; first decide how you would handle both outcomes.
4. The Key to Getting Rich Slowly Is Avoiding One Irreversible Drawdown
巴菲特 recalled his entrepreneurial partner 里克·盖林 by saying that the 3 men were equally intelligent, but 巴菲特 and 芒格 were willing to get rich slowly while 盖林 wanted to “get fat in a single day.” 盖林 later ran into liquidity pressure after levering up during the 70s of the 20th century and had to sell his shares cheaply to 巴菲特.
朱宁 believes that earning an extra 5 points or 10 points in the short term is not that important. What truly drives long-term compounding is avoiding a major drawdown and maintaining an investment philosophy that can be executed across cycles. “Do not lose money and protect the floor” is closer to 巴菲特’s method than forecasting every swing.
People who can predict whether prices will rise or fall tomorrow or 6 months from now are vanishingly rare. 朱宁 jokes that anyone with that ability would not need to invest and could simply buy lottery tickets. Ordinary investors should start from the reality that this ability does not exist, rather than from the illusion that their own judgment is 100% correct.
5. Beta Sets the Starting Point; Alpha Determines Whether Luck Survives
When 朱宁 asked 席勒 what he most wanted to know, 席勒 said he wanted to understand how much of investment success came from luck and how much from ability—the ratio between Beta and Alpha. 朱宁’s choice is clear: if forced to pick one, he would choose luck, because “the broader environment, the era, and the trend” are indispensable.
巴菲特 has also acknowledged that he and 芒格 benefited from the “ovarian dividend”: being born in the United States, being white, receiving an education, and catching the long-term rise of the U.S. economy and stock market. Many people were born in the same era, but only a few became legends, showing that Beta comes first but does not make Alpha irrelevant.
利弗莫尔 made a fortune shorting the market in 1929, then blew up after his confidence rose and he continued increasing leverage, eventually taking his own life. 朱宁 uses the counterexample to make the point that “you can never make money beyond your understanding”; money won through luck can also be returned with interest.
Research on identical twins in the Nordics finds that genes remain the largest influence on investment, saving, and risk-taking decisions, while education and the surrounding environment also have significant effects. 朱宁’s attitude is therefore “do everything humanly possible, then accept what fate decides”: acknowledge innate traits and the era without using them as an excuse to do nothing.
6. Life Has Multiple Waves; Understanding May Mature Only by the 3rd
曲凯 estimates that an individual wave lasts 10-20 years on average. During the 1st wave, a new graduate may simply find themselves inside it; by the 2nd, they may recognize what is happening but lack firsthand experience; only by the 3rd might they have capital, judgment, and the ability to act at the same time. 朱宁 broadly agrees that these capabilities accumulate in stages.
The two left an important tension around graduation timing. 曲凯 cited research showing that people who graduate during recessions may lag in long-term financial outcomes; but a Stanford study of Wall Street found that those who entered at the bottom could be more successful because they genuinely liked the work, had the ability to get in, and were able to wait through the next full wave.
朱宁 lived through the 1997-1998 internet bubble, the 2008 crisis and bankruptcy of Lehman Brothers as a senior executive, and China’s property cycle over the past 10 years. His understanding of investment and wealth also deepened gradually: “Even if you did not make money at the beginning, you may have made experience and ability, and one day they may turn into money.”
巴菲特 made 95% of his wealth after age 65. His principal base was large enough by then, but his understanding had also matured further. 朱宁 consequently asks whether high-net-worth individuals understand “how their money was made”; once they understand its source, they will at least know how it might be lost.
7. Bull Markets Make It Easiest to Mistake the Elevator for Your Own Ability
曲凯 uses an elevator to describe pro-cyclical success: everyone makes different moves inside the elevator but credits themselves for its movement up or down. Economic troughs can also produce great companies because harsh conditions force founders and investors to see more clearly what is actually creating the result.
朱宁 notes that retail sentiment becomes more powerful in bull markets. Accounts may be making money, but returns can still trail a simple index fund by a wide margin; in bear markets, people become more cautious and pay closer attention to their own investment behavior. “Emotion significantly interferes with the level and quality of decision-making.”
The value of starting to manage money early is not limited to compounding. Losing RMB1,000 when monthly income is RMB5,000 hurts enough to make an impression, but it allows someone to learn early lessons that might otherwise require a much larger loss later—a way to buy decision-making experience at lower cost.
8. “This Time Is Different” Cannot Be Falsified, but Position Size Can Change
朱宁 remembers “This time is different” as possibly the 4 most frightening words in investing, a phrase attributed to 邓普顿, founder of Franklin Templeton; after the 2008 financial crisis, one of his Yale classmates also wrote a book with the same title about the crisis. The danger is that both believing this time is different and believing nothing has changed can be wrong.
His qualified view is that every episode is different, but whether those differences are large enough to change the entire valuation framework and human behavior cannot be asserted lightly. “Every time, there is not that much that is different, but every time there are many differences as it develops”—it depends on whether one looks at the process or the outcome.
Since judgment can never be perfectly accurate, the variable investors can control is how much capital to commit and how much risk to take. Investors spend too much time optimizing signals and too little asking what would happen to their lives if the signal—and the resulting position—were wrong.
9. A Small Principal Base Does Not Mean Leverage Should Shorten Your Life
曲凯 asks whether someone with total assets of only RMB5,000 should take more risk, given that even a 20% annualized return would not change their fate. 朱宁 says the impulse is understandable; Chinese households historically favored property in part because it was one of the few investment channels available with leverage.
The first boundary is identifying one’s true risk tolerance. Someone who can go all in, hit zero, feel no fear or depression, and start again may indeed be able to earn money through risk capacity. The problem is that most people “want to take risks but lack the ability or willingness to bear risk.”
If a loss materially lowers quality of life, the risk is not worth taking. “We make money to live, not live to make money.” Wanting to get rich overnight is natural when young, but later in life people may discover that having somewhat more money does not increase the value of life linearly.
People have 2 ways to make money. Financial investment can be fast but volatile; human-capital investment is slower, but relatively more predictable and may also make you a better person. 朱宁, as a professor, is relatively conservative and especially unwilling to see one all-in bet and bad luck permanently damage a long-term life plan.
10. Calling the Top Is Behavioral Finance’s “Holy Grail”—and Prediction Itself Gets in the Way
伯南克 called understanding the Great Depression the holy grail of macroeconomics. 朱宁 proposes the corresponding holy grail of behavioral finance: anticipating the top of a bubble. Behavioral economics recognizes that emotion and error can push prices far from fundamentals, but still has no reliable method for calling the top.
One line of research watches the slope of price gains because “the slope consumes money.” When capital is limited, a rally that moves too quickly and violently becomes difficult to sustain. Quantitative easing in 2008 and again during the COVID-19 pandemic in 2020 created unprecedented liquidity; low rates then pushed investors toward risk, speculation, and leverage, so bubbles became more numerous rather than less.
In 2015, many people expected the stock market to return at least to the previous cycle’s 6,100 level, but it struggled to advance beyond 5,100. 朱宁’s explanation has a game-theoretic element: once a top becomes consensus, someone will exit early, making the publicly predicted top impossible to reach.
The deeper irony is that a bubble can exist only if someone believes it is not a bubble. If everyone decides prices are too high and sells, the bubble disappears on the spot. So “as long as there are still people who believe it is not a bubble,” it may not yet have completed its final stage.
11. When the Last Shorts Capitulate, the Last Buyer May Be Taking the Baton
Tiger Fund had spent the previous 1-2 years shorting the Nasdaq and the internet bubble, but the market kept rising until it was crushed and liquidated in March or April 2000. The Nasdaq peaked 1-2 months after its exit. The fund got the direction right but could not survive the path before the thesis paid off.
朱宁’s mechanism is straightforward: once everyone who believes the market is in a bubble has been crushed by reality, they start saying, “Maybe this is not a bubble after all. I have to buy some too.” The remaining demand is released in a concentrated burst, potentially bringing the bubble close to its top. In crypto and AI, the most determined skeptics flipping from short to long may not be good news.
巴鲁克’s long-term signal was: “When beggars, shoeshine boys, and manicurists are telling you how to make money in the stock market, the market is not far from a top.” 曲凯’s contemporary version is that once something is being widely discussed on Xiaohongshu, its spread may already be well advanced.
12. Your Social Circle Determines Not Only What You Know, but How Many Years Late You Learn It
Research 朱宁 conducted in North America found that individual investors’ main information sources are often not research reports, due diligence, or company visits, but friends, relatives, and social media. Your social circle first determines whether key concepts enter your field of vision, and then whether the information has already been fully reflected in prices by the time it reaches you.
In 2015, people who did not even understand price-to-earnings ratios sold their homes to trade stocks because “4,000 is only the new starting point of the bull market.” AI works the same way: insiders may have positioned 3 years ago, PE investors 2 years ago, and secondary-market institutions last year, while retail investors are only now hearing that AI or optical modules matter. The opportunity may already have been heavily pulled forward by earlier capital.
Early investors still have a low-cost buffer even after a theme becomes crowded. By the time latecomers hear about it, other participants may have completed their allocations long ago. The same information does not have the same value when it arrives at different times; that is the handoff risk embedded in an “information cascade.”
朱宁 relays 席勒’s line from a foreword: investing is both “an activity against human nature” and “a competitive activity.” He also uses 巴菲特’s poker-table analogy: if you sit down and, after a while, still cannot tell who the fool is, “then you are the fool.” Retail investors should at least ask whether their trading counterparties know more than they do.
13. AI Takes Information Equality Toward Capability Equality—and Creates a Double Hallucination
The internet made information theoretically searchable by anyone. AI lets agents search, organize, and analyze on a user’s behalf, which 朱宁 calls the shift from “equality of information” to “equality of capability.” That progression could further accelerate every form of technological evolution.
If an efficient market means prices reflect information more accurately, AI is pushing markets rapidly in that direction. 朱宁 expects more retail investors to access markets through index funds, potentially allowing institutions tied to index ETFs to capture market share even faster.
By the time of the conversation, the market already showed a structure in which indexes were rising while the number of declining stocks exceeded advancing stocks, especially in the United States. Investors who continue trading individual names may not share in the index gains, while institutions retain professional IT, research, organizational systems, and discipline—advantages that do not disappear simply because individuals can call GPT.
AI creates “2 types of hallucination”: the model itself hallucinates, while retail investors may hallucinate that “with GPT, I am not really different from an institution.” When 朱宁 asked GPT and DeepSeek what a novice should do, the answer was “If you can’t beat them, join them”—become an ETF or fund investor.
14. The Problem with Value Investing Is Using the Short Term as the Scoreboard
If value investing means buying an asset below the value of the underlying business, 朱宁 says it “can never be wrong and can never fail.” The prerequisites are genuinely understanding business value and having 巴菲特-like patience and capital to hold for the long term.
What really changes are the benchmark and the time horizon. Should performance be measured against the Nasdaq over 6 months, 2 years, 20 years, or 60 years? On almost any individual 2-year window, 巴菲特 has had long periods of underperformance; over 60 years, compounding becomes the legend.
During the 1998-2000 internet bubble and the 2008 financial crisis, 巴菲特’s value investing was mocked, only to “have the last laugh” both times in hindsight. 朱宁’s conclusion: “Investing is a marathon, not a 100-meter sprint. It is not about who laughs first, but who laughs longest.”
15. AI May First Replace Its Own Creators, but the Employment Shock Will Not Unfold in a Straight Line
朱宁 believes AI-based disruptive applications will materially displace jobs in the future, affecting both white- and blue-collar workers. For now, the people most quickly exposed may be AI’s inventors and programmers. As for whether AI could develop consciousness, surpass humans, or even turn the sci-fi “killer network” into reality, he says “it is not impossible.”
Governance must ultimately return to compassion, empathy, and goodwill toward one’s fellow humans and animals. Those capabilities must either be transferred to AI or protected by governance institutions. Mass unemployment could become reality quickly, making UBI a necessary topic of discussion, but 朱宁 does not believe it would automatically produce an overly optimistic communist vision.
He also rejects simplistic alarmism. The steam engine, the internet, and the “millennium bug” all generated widespread fear, while U.S. employment data released before the conversation proved unexpectedly strong and resilient, even changing market expectations for Federal Reserve rate cuts. Projecting next year’s AI landscape in a straight line from past experience may itself be the mistake.
The response has 3 parts: embrace the technology, promote intergovernmental and international cooperation on governance, and use arrangements such as UBI to free individuals to pursue their goals. AI may make basic-skill training less important and allow people who cannot draw to create their own art; beyond the disruption, “many wonderful things may happen.”
16. Policy Can Smooth the Cycle, but It May Simply Swap One Bubble for Another
朱宁 uses counterfactual history to argue that even without a particular president or specific event, the Great Depression might still have unfolded along broadly similar lines, because policy itself reflected the economic realities and social mood of the time. “What is coming will still come”; individual policies mainly change the path and timing.
The Federal Reserve was able to quickly reduce the shock after the internet bubble burst, but 克鲁格曼 criticized it for blowing up a housing bubble to pay for the internet bubble. Then, in response to the housing bubble, another internet or AI bubble could emerge.
朱宁 still believes the market has powerful vitality. Governments can distort, control, and smooth short-term cycles, but they have difficulty eliminating long-term trends. Assuming the Fed does not adopt extreme policies, he believes current U.S. equity valuations will be difficult to sustain, although the process could last a long time.
AI does have profitable companies, giving it an additional layer of support compared with many unprofitable internet companies in 2000. But making money and deserving today’s valuation are entirely different questions. Finance ultimately comes down to whether the numbers work; interest rates and funding tightness could at some point become the last straw.
17. Narratives Can Turn Capital into Technology, but They Cannot Make a Tree Grow into the Sky
From the private market, 曲凯 sees new concepts quickly forming consensus and absorbing capital as old narratives cool. Enough money can even compress a technology cycle that would normally take 5 or 10 years, validating the narrative in return and creating a feedback loop between capital and progress.
朱宁 believes capital can accelerate technological progress, but insists that technology remains primary: “You can fertilize and water a tree more heavily and make it grow faster than before, but it cannot grow into the sky.” On humanoid robots, he raises a basic question: humans did not rule the world because they evolved into a human shape, so why must robots look human?
The positive function of a bubble is to accelerate investment cycles and technological progress dramatically. The internet buildout of 1998-2000 produced enormous waste, but it left infrastructure and experience for Web 2.0, Web 3.0, and eventually AI. A technology’s long-term value can be real without preventing a period of bursting and elimination in between.
A narrative mobilizes not only money, but also talent and society-wide resource allocation; the choice of majors by students at Tsinghua University and Shanghai Jiao Tong University offers a window into that process. Pure business models such as O2O can be fleeting, while AI has produced real progress at the technology, application, and social levels. The narrative is therefore “adding momentum, not conjuring something from nothing.”
18. Four Conditions Can Define the Bubble Zone, but Cannot Declare an Imminent Collapse
朱宁 wrote Rigid Bubbles in 2016 and published a new edition in 2020. By 2021, he still maintained that only hindsight can prove whether a judgment was correct. Bubble research can identify a high-risk combination, but it cannot provide certainty in advance.
The 4 “necessary but insufficient conditions” are a new concept, product, or technology; very loose liquidity; government support; and inexperienced, typically younger investors willing to put real money behind the future. A bubble will almost certainly contain these conditions, but having all 4 does not guarantee a crash.
Electronic advances—including vacuum tubes and transistors—in the 60s and 70s of the 20th century offer one example. The sector rallied sharply and then pulled back, but it did not halve or lose as much as two-thirds as the internet sector later did; it subsequently resumed rising with the “Nifty 50,” before a major market adjustment accompanied the Middle East oil crisis in the early 70s.
A decline is inevitable; its magnitude determines what it gets called afterward. 朱宁 acknowledges that this is not a scientific threshold, but if forced to choose, he would treat a 40%-50% decline as a bubble, while 20% would look more like a healthy correction.
朱宁 also recalled that, around 2018, 席勒’s valuation framework put U.S. equities second only to 2000 in expensiveness; he separately said current U.S. equity valuations are about 20% above 1929 levels.
19. Recognizing Bias Does Not Create Immunity, and Wealth Is Not Life’s Final Score
朱宁 admits that knowing is easier than doing. The day before Lehman’s bankruptcy, he was holding a housewarming party in Hong Kong; among the 30-40 guests, many were Lehman colleagues, yet no one said the company would be bankrupt the next day. His own career choice had also involved overconfidence, insufficient due diligence, and an inadequate understanding of the world.
Financial literacy should work like a visual correction for the near-large, far-small illusion. When a distant building looks small, you understand that the effect is caused by distance; when a pie appears in the sky, you should likewise recognize that “it will not fall into my lap.” There must be risks you do not yet know.
Wealth is only one part of life. No matter how much money you have, you still eat 3 meals a day and sleep in 1 room; “with greater ability comes greater responsibility,” and money is also a form of responsibility that not everyone can handle. Research on lottery winners shows that enormous wealth did not improve quality of life and instead often created new problems.
If he could return to his youth, 朱宁 would try more possibilities, because “the greatest value of being young is that you can fall down and start again,” and he would be more faithful to his inner convictions and do more meaningful things. 卡尼曼 distinguished between happiness experienced in the moment and happiness remembered afterward; the 2 can conflict. So do not compare yourself horizontally—just live your own journey.